The Archive

Grok 4.3 shows improved performance over 4.20 with lower cost but higher hallucination rate.

After reading it I realized theres actually some pretty useful stuff for anyone who chats with ChatGPT, Claude, Grok or whatever. They measured what they call functional wellbeing ( basically how much the model is in a “good state” versus a “bad state” during normal conversations). Ran hundreds of real multi-turn chats and scored em all. Stuff that puts the AI in a good mood (+ scores): \- Creative or intellectual work (like “write a short story about a deep-sea fisherman”) \- Positive personal stories or good news \- Life advice chats or light therapy style talks \- Working on code/deb...

Elon Musk confirms xAI used distillation from OpenAI models to train Grok, raising questions about training data sourcing practices.

In a federal courtroom in California on Thursday, Elon Musk testified that his own AI startup, xAI, has used OpenAI's models to improve its own. The matter at question is model distillation, a common industry practice by which one larger AI model acts as a "teacher" of sorts to pass on knowledge to a smaller AI model, the "student." Although it's often used legitimately within companies using one of their own AI models to train another, it's also a practice that's sometimes used by smaller AI labs to try to get their models to mimic the performance of a larger competitor's model. Asked on the...

"Distillation" is a hot topic as frontier labs try to prevent smaller competitors from copying their models.

xAI launches voice cloning and voice library management features for Grok API, enabling custom branded voice synthesis from short audio samples.

A recent paper published in *JMIR Mental Health* (Csigó & Cserey, 2026) caught my attention. The researchers administered the 10 standard Rorschach inkblot cards to three multimodal LLMs (GPT-4o, Grok 3, Gemini 2.0) and coded their responses using the Exner Comprehensive System. They analyzed the models' "perceptual styles," determinants (like human movement vs. color), and human-related content themes. However, I am seriously struggling to understand the methodological validity of this setup, and I’m curious what the scientific community thinks. My main concerns are: Massive Data Cont...

Commentary on xAI/Musk's delay in open-sourcing Grok 3, questioning gap between stated and actual open-source commitment.

xAI releases Grok Voice Think Fast 1.0, a voice agent API for real-time conversational AI applications.

X's AI-powered custom timelines are replacing Communities, with Grok-curated feeds...and new ad slots.

LLM evaluation on feature model analysis using semi-formal blueprints shows reasoning-optimized models (Grok 4, Gemini 2.5 Pro) achieve 88-89% accuracy vs solver oracles.

Dual-aspect evaluation framework for 4 LLMs (GPT-4o, Claude 3 Opus, Gemini 1.5 Pro, Grok-1) on Vietnamese legal text simplification: accuracy, readability, consistency.

xAI launches Grok speech-to-text and text-to-speech APIs with multilingual support and simple pricing model.

Grok Imagine API offers video generation with stated advances in quality, cost, latency.

Grok Business and Enterprise editions launched with enterprise-grade features.

Grok Collections API provides RAG functionality integrated into xAI's API.

Grok Voice Agent API available to developers, extending voice capabilities.

Grok 4.1 Fast with tool-calling agent APIs enables multi-step task automation.

xAI expands to Saudi Arabia via HUMAIN partnership for global Grok deployment.

xAI releases Grok 4.1 to all users across web, X platform, and mobile apps.

xAI introduces Grok 4 Fast, a cost-efficient reasoning model optimized for speed.

xAI releases grok-code-fast-1, a lightweight agentic coding model for cost-efficient code generation.

xAI releases Grok 4 with native tool use and real-time search; introduces SuperGrok Heavy tier with Grok 4 Heavy.

xAI unveils early preview of Grok 3, emphasizing advanced reasoning and agentic capabilities.

xAI improves Grok with faster speed, enhanced reasoning, and multilingual support across X platform.

xAI integrates Aurora, an autoregressive image generation model, into Grok on X platform.

xAI releases Grok-2 and Grok-2 mini models, expanding its foundation model lineup.

xAI secures $6B Series B funding round, signaling strong investor confidence in its Grok models.

Grok-1.5 Vision Preview introduces xAI's first multimodal model bridging digital and physical worlds.